MLOps Engineer

Dex; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

Dex is seeking an experienced MLOps engineer to own deployment patterns and observability for LLM services within sovereign Kubernetes environments in London. You will work between distributed systems engineering, GPU scheduling, and agent execution logic to ensure reliability under enterprise loads.

You will design scalable deployment patterns, optimize costs, and implement trace-level observability across complex multi-step agent workflows, enabling robust AI platforms in regulated settings.

Qualifications

• 3+ years in MLOps/LLMOps or AI platform engineering.
• Production experience with containerised workloads on Kubernetes.
• Strong Python and IaC/CI-CD automation.

Responsibilities

• Architect and automate zero-downtime deployment pipelines for LLM services on Kubernetes.
• Build end-to-end evaluation and observability to trace agent execution and model drift.
• Own GPU/CPU allocation, cluster autoscaling, and compute efficiency.
• Maintain stability and latency of real-time data subsystems and feature stores.
• Diagnose production failures spanning agent runtime, orchestration, and cloud infra.

Skills

Python

Kubernetes

CI/CD

IaC

Observability

Tools

OpenTelemetry

Prometheus

Grafana

LangSmith

Langfuse

Arize

Job description

This role is with one of Dex's trusted partner companies. We work closely with their teams to truly understand their culture, goals, and what they're looking for, so we can match you with the right opportunity and give you context about the role before you commit to a process.

The role

Most enterprise AI rollouts rely on public cloud APIs that heavily regulated organisations cannot touch. This London-based infrastructure company builds sovereign runtime environments that allow defence, intelligence, and regulated enterprise teams to deploy frontier and open-weight models entirely inside their own perimeters, including air-gapped clusters. Backed by over $70M from tier-one global investors and led by alumni from top-tier enterprise software and defence engineering, they are tackling the foundational reliability problem of sovereign AI.

As the first dedicated MLOps engineer on the team, you will own how complex LLM services and multi-step agentic applications run, scale, and recover on high-throughput Kubernetes infrastructure. This is not a maintenance seat running standard CI pipelines or an ML research role isolated in notebooks; you will sit directly between distributed systems engineering, GPU scheduling, and agent execution logic. You will define the deployment patterns, cost optimizations, and trace-level observability that keep non-deterministic systems reliable under real enterprise loads.

The work

• Architect and automate zero-downtime deployment pipelines for LLM services and autonomous agent workflows on Kubernetes, using canary and blue/green rollout strategies.
• Build the end-to-end evaluation and observability framework to trace non-deterministic agent execution paths, monitor hallucination rates, and detect model drift.
• Own GPU and CPU allocation, cluster autoscaling, and compute efficiency to drive down time-to-first-token while controlling infrastructure expenditure.
• Maintain the stability and latency of critical data subsystems, including real-time vector indexing, embedding pipelines, and feature stores.
• Diagnose complex production failure modes that span agent runtime logic, orchestration frameworks, and underlying cloud infrastructure.

What You Bring

• 3+ years in MLOps, LLMOps, or AI platform engineering, with proven experience running containerised workloads on Kubernetes at production scale.
• Strong Python engineering ability paired with practical fluency in Infrastructure as Code and modern CI/CD automation.
• Hands-on experience instrumenting observability and LLM tracing across distributed systems using tools like OpenTelemetry, Prometheus, Grafana, LangSmith, Langfuse, or Arize.
• Operational familiarity with agentic architectures, including state persistence, tool execution boundaries, and failure handling across multi-step agent runs.

As part of the recruitment process at Dex, we process your personal data in accordance with our Privacy Notice for Job Applicants. This notice explains how and why your data is collected and used, and how you can contact us if you have any concerns.